Dysphonic Voice detection using MDVP Parameters

Abstract

A new noninvasive method is introduced to detect dysphonic voice using Multi-Dimensional Voice Parameter (MDVP) parameters. Voice can be considered as a multidimensional measurable event. Dysphonia can be accounted as the first presenting symptom of larynx cancer and is caused due to defective mucosal vibrations. MDVP is a commercial software which is very useful in the acoustic analysis of a given voice. Here sustained vowel /a/ is used and 30 MDVP parameters are used for the analysis. This describes the voice objectively and the variation of the parameters can be accounted as the indication of some abnormality. These parameters are fed to an Artificial Neural Network (ANN) which classifies pathological voice from healthy voice. The system gives an overall efficiency of 93.33%.